Converting Markdown to ePub or Mobi Using Pandoc

Converting Markdown to ePub or Mobi Using Pandoc(themythicalengineer.com)

128 points by sks147 5 years ago | 54 comments

d_rc 5 years ago |

Thanks for the nice tutorial OP.

Here's a public notebook in Deepnote if anyone wants to play around with the code or duplicate it: https://deepnote.com/project/Converting-Markdown-to-Epub-or-...

2 fun facts about Deepnote:

1. You can create a Custom environment by writing a Dockerfile with all the libraries you need to install and everytime you're in a need to re-use a similar functionality (e.g. convert yet another book to mobi), you can just fire it up and all will be preinstalled. https://docs.deepnote.com/environment/custom-environments

2. You can turn any notebook to a blogpost right away and publish within Deepnote directly.

Disclaimer: I'm a software engineer at Deepnote.

OhHiMarkos 5 years ago | |

Wait this is a notebook similar to pythons' notebook but it's a docker environment where I can install a lot of stuff I want and then do even more stuff? Am i getting this right?

It's like a shell to a vm but in a notebook format that you can then use to blog?

d_rc 5 years ago | | |

Yeah, you got it right. You can even access the actual shell in the vm, not just the notebook environment.

sireat 5 years ago | |

what would be advantages to going to Deepnote from regular Jupyter notebooks based workflow?

Let's assume someone who has been working with Jupyter notebooks(mostly Python based) for a long time.

Are Deepnote notebooks exportable?

The big worry is that you guys decide to pivot or radically change your pricing model and there is no offramp.

By comparison I don't mind using Google Colab. If Google Colab decides to shutdown or 100x their price I can take my .ipynb files and use them on my local littlest JupyterHub instance.

d_rc 5 years ago | | |

Deepnote internally supports .ipynb format and you can always export the Deepnote notebook to .ipynb similarly as you'd in Colab.

In general the main selling points are live collaboration (you can work on a notebook with you team as you'd do on a google doc), and integrations (you can plug-in your snowflake db, or s3 bucket or whatever, and have it connected for any further analysis, or a long-term training, etc.

For many non-software-developer data scientists, it's also easier to work in a cloud environment compared to installing stuff locally, and to version their notebooks in Deepnote instead of git. But this really depends on the particular workflow that one has.

sks147 5 years ago | |

Thanks for publishing this as a notebook. I really love the platform.

ggambetta 5 years ago |

I followed a similar approach for my novel; started with Markdown, used pandoc to convert it to epub/mobi, but also to LibreOffice .odt to generate the PDF for the paperback. Wrote some details about the process here: https://gabrielgambetta.com/tgs-open-source.html

PascalPrecht 5 years ago |

That's how I wrote and self-published my book as well! Although, I created a script that turns md to epub/mobi/pdf using pandoc.

Here's how I did it in case anyone is interested: https://pascalprecht.github.io/posts/writing-an-ebook

sks147 5 years ago | |

This is great. Thanks for sharing

kimi 5 years ago |

Don't want to be a troll, but if you are writing anything that is not a README and/or is a book or booklet or bookish, do yourself a favor and use Asciidoc instead.

debiandev 5 years ago | |

This! Asciidoc is the grown-up brother of Markdown. Designed to scale up to entire books, handle images, tables, references, citation, book indexes, maths.

And the syntax is very friendly and intuitive.

kimi 5 years ago | | |

And it's quite easy to insert compiled images, e.g. Graphviz, UML diagrams, Ditaa, just by having the SOURCE in your document.

davegauer 5 years ago | |

Absolutely. AsciiDoc (2002) is actually 2 years older than Markdown (2004), but is surprisingly similar to write.

It was created as an equivalent to DocBook XML for the creation of book-length technical documents. It has a rich history and is well supported in many places (try writing a README.adoc for your next GitHub-hosted repo).

It is also currently undergoing a standardization process:

https://projects.eclipse.org/proposals/asciidoc-language

dredmorbius 5 years ago | |

What's your suggested Asciidoc toolchain?

(Pandoc will ingest Asciidoc as well as Markdown, FWIW.)

kimi 5 years ago | | |

At work, we have been using Asciisoc for 10 years or so, so we use Asciidoc -> XSLT -> PDF / chunked HTML. Works great, not nice to use.

For new projects, we use Asciidoctor - it does everything but chunked HTML, and it's a pleasure to use.

jtbayly 5 years ago | |

Give me a reason to use asciidoc rather than bookdown.

gexla 5 years ago | | |

I assume the person you're replying to was referring to Markdown rather than Bookdown. It seems that Asciidoc was designed for the direction which Markdown has going with all these variants. If you find yourself chasing these Markdown variants to get more flexibility, then Asciidoc might be what you're looking for.

I don't know anything about Bookdown and it may be similar to Asciidoc. I would be willing to bet that Asciidoc would be around longer than most of the MD variants though.

Why might you want to continue using MD flavors? You already have loads of MD docs and you have no control over the processes which create them (shared environment, higher-ups force you to use MD, etc.)

NOTE: If you're interested in Asciidoc, also take a look at Asciidoctor.

mraza007 5 years ago |

Pandoc has been one of the best tools I have used and this blogpost is well written

avinassh 5 years ago |

This is a nice tutorial, thanks for submitting it! However, for me, the biggest discovery was epub.press [0]. I just tried for couple of open pages, it works quite well!

[0] - https://epub.press/#about

progval 5 years ago |

Why wget|dpkg and wget|sh instead of apt to download Pandoc and Calibre?

You should be able to replace all this:

    !wget https://github.com/jgm/pandoc/releases/download/2.11.3.2/pandoc-2.11.3.2-1-amd64.deb
    !sudo dpkg -i pandoc-2.11.3.2-1-amd64.deb
    !apt install libgl1-mesa-glx -y
    !wget -q -O- https://download.calibre-ebook.com/linux-installer.sh | sudo sh /dev/stdin

With simply this:

    !apt install pandoc calibre

sundarurfriend 5 years ago | |

Calibre website strongly recommends downloading from their site instead of OS packages, mentioning that the packages are often out of date. And I've generally found this to be true - Calibre versions on package repos are often several versions behind, more than the usual "package maintainer trying to play catch up" differences.

I'm usually averse to the wget|sh installs, but in this case it seems worth it. You can inspect the .sh file (which is really mostly Python code) before running it, just to not get into the bad habit of directly piping in code from the internet.

progval 5 years ago | | |

> You can inspect the .sh file

That's not the issue. Installing software this way means you have no automatic updates in the future. It's fine if you re-run the install script on a regular basis (eg. by recreating containers) AND you don't pin versions

But OP's instructions fail both: there is no mention of updates, and they pin the pandoc version.

stonesweep 5 years ago | |

I use pandoc in a CD pipeline, the version in the repos is stale compared to upstream (normal, that's how it is) unless you're on a rolling distro like Arch.

I have reported pandoc bugs and had them fixed (great dev team), pulling the latest single-DEB install (no deps, unlike the one in the Debian repo) and using it gets all the latest updates which matter to a process like this.

In this particular case your needs to use the latest pandoc lead to the wget pull and install, which thanks to their DEB design is easy and clean to do in an ephemeral CI container.

coreypreston 5 years ago | | |

Do you have any more details about how you integrated Pandoc into your pipeline? A post or something?

asicsp 5 years ago |

The tutorial is presented well. My biggest takeaway was that one can use 'Deepnote' to run Linux commands.

If you are interested in knowing how to customize `pandoc` for generating PDF/EPUB, I have a tutorial [0] based on books I've written. I also have links at the end with related resources, including tools others than `pandoc`.

[0] https://learnbyexample.github.io/customizing-pandoc/

throw14082020 5 years ago | |

Jupyter notebook also has the same feature. `!shell command` or start a cell with `%%bash` and everything in it will run through the terminal, not the notebook interpreter.

asicsp 5 years ago | | |

That'd require you to have access to a *nix terminal on your system. Deepnote is allowing access through their servers, so for example, you can try this tutorial on Windows.

See this comment from Deepnote engineer for more details: https://news.ycombinator.com/item?id=26899905

Finnucane 5 years ago |

Note that mobi format is being deprecated by Amazon. If you're producing a file for distribution by Amazon, you only need the epub file.

Turing_Machine 5 years ago | |

However, Kindles can't read epub directly. If you're trying to produce a file you can sideload on your (or your customer's) Kindle without going through Amazon, you're still gonna need mobi, or one of the later proprietary (and undocumented) Amazon mobi variants.

Finnucane 5 years ago | | |

True, but I wouldn't count on that being supported forever.

alias_neo 5 years ago |

Is Pandoc being used mostly to join the files?

I recently started converting Markdown files to epub (and kepub) for my new Kobo. I load the Markdown straight into Calibre though.

On a side note, is there some benefit to mobi over epub? Kepub seems to be the preferred format on Kobo, because for some reason it turns pages _much_ faster than epub and gives access to reader statistics (if one cares about that).

pronoiac 5 years ago | |

Hey, I'm curious about your Calibre usage! I'm working on turning a written book to Markdown, and pandoc has a real pain point on links between chapters. Please tell me more!

alias_neo 5 years ago | | |

What would you like to know? In Calibre you can set regex to indicate what should be used for chapter, sub chapter etc, and it can be used to generate the TOC. So I use Markdown headings, #, ##, ### etc for chapters and subsections.

hoppyhoppy2 5 years ago | |

>On a side note, is there some benefit to mobi over epub?

AFAIK Amazon Kindle devices can read mobi files but not epub files (unless you convert them to something else first).

(The mobi format is older, so if you want to read an ebook on your old Palm Pilot PDA then you'll probably want mobi.)

alias_neo 5 years ago | | |

Thanks for the heads up! As for my palm pilot, well, I'm not sure which drawer that's been in for the last decade or two.

acidburnNSA 5 years ago |

Good stuff. When I wrote an eBook I found the extra features of reStructuredText to be useful (index, glossary, graphviz & Tikz environments, etc.) and wrote a sort of similar post.

https://digitalsuperpowers.com/blog/2019-02-16-publishing-eb...

masklinn 5 years ago | |

Yeah the rSt markup is pretty idiosyncratic and both anal-retentive and a bit inconsistent (e.g. it can be hard to internalise where it does or does not want blank lines) but Sphinx is an actual document system, having to work with markdown for anything beyond a single file quickly gets painful.

throwanem 5 years ago |

You know, I bet it wouldn't take that much to go from epub to PDF, suitable for printing and binding. The structural information is pretty much all there, I think - it'd just need pagination and formatting for print, really.

I'd definitely want to use such a thing, as a way to feed my bookbinding hobby. I wonder if anyone else would?

timClicks 5 years ago |

Is the resulting typesetting any good?

maxnoe 5 years ago | |

Epub is just html with css and images in a zip file. So the quality of the typography will largely depend on the renderer, not the file itself.

fau 5 years ago | | |

Unfortunately, the renderer that gets used will sometimes itself depend on the file. For example, if you transfer the generated Mobi file to a Kindle as the article says, it will get rendered using an inferior renderer (in terms of kerning, for example) compared to the renderer that would've been used had the file been a KFX ("Kindle Format 10").

dhruvdh 5 years ago | | |

The end result would largely depend on the quality of generated HTML and CSS, no? That's what the OP is most likely asking about.

5tefan 5 years ago | |

I would like to have better type setting. It is a pleasure to have a nicely layouted and set page. My layman's take would be that the rendering engines have headroom to improve but most readers don't really care.

nvr219 5 years ago |

Pandoc is a GOAT tool! It's so good.

TTT1 5 years ago |

very interesting